[test-qa-hourly] test: lock past relative-time boundary

[i-xtsu-sixyou-ken-mei]

Intent

Continue continuous QA with one small regression-focused Control Center test for the past relative-time threshold.

Inspection

• Latest main: 42c3bdb7a811a596aec68ddec977a4038437bb7e (merge of PR fix(control-center): load Supabase env for direct local dashboard startup #352).
• Previous hourly QA PR [test-qa-hourly] test: lock future relative-time boundary #363 is merged; no reusable open PR matched [test-qa-hourly], label test-qa, or branch prefix test-qa/ongoing-.
• Open PRs reviewed: fix(podcast): use canonical video poster for YouTube thumbnails #360, fix(podcast): simplify visual subject search and expose debug plan #362, fix: recover stalled podcast ingest with paid translation fallback #364. None changes apps/control-center/src/client/format.test.ts; fix(podcast): simplify visual subject search and expose debug plan #362/fix: recover stalled podcast ingest with paid translation fallback #364 touch other Control Center podcast surfaces only.
• Latest main did not expose a deterministic red core QA gate during inspection. The preceding PR head had active/successful checks and no observed failure requiring a CI-first repair.
• relativeTime() uses relative minute labels for past timestamps strictly under one hour and switches to absolute locale time at exactly one hour.
• Future 59/60-minute behavior is already locked by the previous QA PR, but the symmetric past boundary was not covered.
• .github/workflows/ci.yml still runs quick gates (format repo contracts), code quality (type-check lint deadcode dup), tests (test analytics), E2E, security, and deploy gates unchanged.
• Coverage aggregation remains scripts/coverage-summary.ts, supporting Istanbul/V8 summaries and pytest Cobertura input.

QA change

Strengthen the existing recent timestamp test with two boundary assertions:

• 59 min ago remains relative (59 min ago).
• exactly 60 min ago switches to the absolute locale timestamp contract.

This is test-only. No implementation, workflow, CI gate, threshold, ignore, docs, README, CLAUDE, or agent instruction changed.

Validation

• GitHub compare main...test-qa/ongoing-coverage-20260903 — PASS: 1 commit ahead, 0 behind; exactly one changed file; +4/-0.
• Static implementation/test contract review — PASS: assertions match the current strict < 60 * 60_000 past-time branch.
• Focused local test command — NOT RUN in this connector-only automation runtime; no repository execution shell is available here.
• Broader format/typecheck/lint/test/coverage validation — delegated to this PR's normal GitHub Actions gates and will be treated as authoritative executable validation.

Scope guard

Changed file only:

• apps/control-center/src/client/format.test.ts

No production or documentation changes.

Next QA target

After CI is green, inspect another high-value formatter boundary rather than extending this timestamp family mechanically; prioritize nullable/zero-value formatting or duration rounding only if current behavior is under-tested and user-visible.

[i-xtsu-sixyou-ken-mei]

Hourly QA update — CI-first investigation

Inspection

• Reused this open QA PR: [test-qa-hourly], test-qa, branch test-qa/ongoing-coverage-20260903.
• Latest main remains 42c3bdb7a811a596aec68ddec977a4038437bb7e; this PR is still a single test-only change in apps/control-center/src/client/format.test.ts (+4/-0).
• Reviewed active PR scopes (fix(podcast): use canonical video poster for YouTube thumbnails #360, fix(podcast): simplify visual subject search and expose debug plan #362, fix: recover stalled podcast ingest with paid translation fallback #364). fix: recover stalled podcast ingest with paid translation fallback #364 already carries repository-wide formatting cleanup for the two main-tree files identified below as part of its larger ingest-recovery branch.
• Re-read .github/workflows/ci.yml: quick=format repo contracts; quality=type-check lint deadcode dup; tests=test analytics; E2E/security/deploy gates unchanged.
• Re-read scripts/coverage-summary.ts: coverage discovery/aggregation still supports Istanbul/V8 and pytest Cobertura workspace reports.

CI #2112 / run 33708314886

Result: failure, isolated to quick-gates.

Successful jobs observed: tests, coverage, code-quality, security, e2e, deploy-gates, check-dead-env.

Downloaded and inspected artifact quick-gates-verify-1:

• repo — PASS
• contracts — PASS
• format — FAIL

format.log shows the formatter changed only two files, both already present on main and unrelated to this PR's test:

1. apps/control-center/README.md — adds the missing EOF newline.
2. apps/control-center/src/server/launch-contract.test.ts — canonical Prettier wrapping of one expect(...).toContain(...) assertion.

The PR's own changed file apps/control-center/src/client/format.test.ts was reported unchanged by Prettier.

Action

No commit added this run. The CI failure is caused by existing main-tree formatting drift, and fixing the README here would violate this QA task's explicit scope guard (do not change README/docs). Adding only the allowed test-file formatting fix would still leave the same quick-gates failure, so that would be noise rather than a coherent CI repair.

Commands / checks

• GitHub Actions run lookup for head fed305186eb32b5ad4a202a870f730ac01b186bf → CI #2112 completed failure.
• Actions jobs inspection → only quick-gates failed; all substantive test/coverage/quality/security/E2E/deploy jobs above passed.
• Downloaded quick-gates-verify-1 artifact and inspected result.json, logs/format.log, logs/repo.log, logs/contracts.log → failure isolated to format; two unrelated main-tree files listed above.
• PR patch review → only apps/control-center/src/client/format.test.ts, +4/-0.

Auto-merge

Not enabled. Blocker: required CI is red. I will not bypass it or broaden this QA PR into a README cleanup.

Next QA target

Once the unrelated main formatting drift is fixed/merged and this branch is synced, rerun CI first. If green, move away from timestamp micro-boundaries and inspect a higher-value user-visible formatter/nullable-zero behavior or another regression-prone Control Center path with weak coverage.